Skip to content

Dataset · multimodalart

Synthetic faces with landmarks and captions

multimodalart/facesyntheticsspigacaptioned

Between 100K and 1M synthetic portraits from Microsoft FaceSynthetics with SPIGA facial landmark annotations and BLIP captions.

What's inside

This is an extended version of the Microsoft FaceSynthetics dataset, including 100,000 to 1 million generated face images and text descriptions. The original pictures are augmented with facial landmark coordinates calculated by the SPIGA model, as well as automatic text captions generated by the BLIP neural network.

What you can do with it

  • Training LoRA and ControlNet for portrait generation using landmarks
  • Training face landmark detection models
  • Fine-tuning multimodal image captioning models
Size
100K–1M rows
Downloads this month
1.5K
Likes
34

How to load it

from datasets import load_dataset

ds = load_dataset("multimodalart/facesyntheticsspigacaptioned")